Problems with new setup

I'm kinda at wits end with my new machine, so I'm turning to you guys for help.

Back in January, I built a new machine for myself, and it ran great. Then Unreal Tournament 2k4 comes out, and I can't play it for longer than 3 minutes. Battlefield Vietnam is horribly unstable. The machine doesn't 'feel' as fast as it should, though benchmarks have it cranking right along with no worries.

Someone suggested that I should get a new PSU, as that may take care of my problems. I was full intending on doing this when on Wednesday my machine took about 7 power hits then a full blackout. When the power was restored, my machine wouldn't even boot. So I get a new power supply, dick around with everything, and then fire the machine back up. Same problems across the board with the added bonus of I have apps now that hang/are really slow to respond. Generic stuff like compilers, Office and even IE.

I then started with the diagnostics:

Memtest86: 3.5 hours for each DIMM, 9.75 hours for both DIMMs, no errors
CPU Stability Test: 8 hours, no errors
Prime95 ( suggested by the Unreal forums ): Fails with an error within 3 minutes
Sandra/3DMark : Complete normally
No Virii/Trojans/Spyware/whatevers found from 3 seperate scans from 3 different prgorams. Nothing odd running in services/task manager.
I haven't yet because I honestly don't think that has anything to do with my problems. Though I may be wrong

Hottest the machine has been (during the CPU Stability Test )
CPU: 62.5C Case: 40C


My specs are:
Generic Case
Antec TruPower 550W PSU ( BRAND new as of Weds )
Belkin UPS ( no more power hits for me )
Abit NF7-S mobo
AMD 2800+
1 Gig Centron DDR400 PC2700 RAM
ATI Radeon 9800 AiW
30 Gig ATA 133 HD ( system drive )
80 Gig SATA Maxtor HD ( game/apps drive )
200 Gig ATA 100 HD ( storage drive )
LiteOn DVD-RW
Windows XP Pro

Everything is running the latest and greatest patches/BIOS's/firmware.

Problems are:
General machine slowness
UT/BFV crashes constantly, making them unplayable
Apps randomly hang/uber slow to respond
Won't pass Prime95

I'm kinda thinking at this point that I have a MoBo problem, but I've seen Abit boards take shit loads of punishment and keep on tickin. Is this Prime95 thing a decent torture test? Anyone have any ideas?

Yes, Prime95 is a good indication of system stability. I believe I have the same config.

Barton 2800+ and Abit NF7-S.

Your CPU seems to be pretty hot. Mine never exceeds 55C. Are you sure the heatsink is mounted properly and the thermal paste is applied properly? If you're using a crappy thermal pad, get rid of it and use Arctic Silver.

Also, what is the CPU voltage at? I don't mean what you have it set for, but what is it ACTUALLY running at according to motherboard monitor? Pay attention to the voltage reading both while the CPU is idle, and then run something like Prime95 or CPUBurn (http://users.ev1.net/~redelm/) and watch the voltage fluctuation. I recommend CPUBurn to do this part of the test.

I had a problem with Prime95 failing at the default CPU voltage, until I noticed that at the low point under 100% CPU load, the voltage dipped down a bit for some strange reason. We SHOULD be at 1.65V minimum all the same, but I noticed it dipped as low as 1.6V. I upped it in my BIOS from the default 1.65 to 1.675 to compensate and reran Prime95 and it passed.

I knew you wouldn't let me down, iNVAR.

My high CPU temp was during the preheat 5 minute stress test of CPUStability. Usually my CPU idles around 45ish, and get's up to around 57 or so under a decent load ( gaming, heavy work stuff, etc )

I tried CPUBurn, which lasted all of about 11 minutes before the sucker crashed my machine. I was keeping an eye on the voltage as you suggest, an the maximum I saw was 1.65, with it going between 1.6 to 1.62. I'll give your suggestion of upping it in the bios a shot and see what happens then.

i still think your system is running extremely hot though. is that the retail HSF you have on there? i took the wimpy fan off (kept the HS) and replaced it with a slightly bigger fan that was audibly still acceptable. it blows about 2x the amount of air on and is still pretty quiet.

before you up the voltage, i'd try to lower that temperature.

Yeah, it's the retail stuff. I'll go grab a new HSF and see what happens.

I bumped the core up to 1.675, and ran Prime95 again. Failed at the same old spot it always does. I watched the voltage and it started out fine at 1.65 to 1.66, and then about 2 minutes into the test it dropped back down to 1.6 to 1.62. Then if failed.

Right now after the test my CPU is running at 43.5C and the case is a balmy 25C


So you're leaning more towards I'm running into some sort of heat issue? Or is this just something you think needs to be rectified? I'm really wondering about the stability of the mobo at this point, or am I just being paranoid?

"Prime95 ( suggested by the Unreal forums ): Fails with an error within 3 minutes"
:O
thats really bad man, but at least you ran the benchmarks to check it out. If you are O/Cing your pc stop and set it back to factory defaults, I would consider maybe even rolling back the FSB from the factroy defaults if it gives you the option in bios just to check it out and see if you can stop the error. Rolling back the fsb will cool the processor a little more too (you are wayyyyyy too hot). If the system isn't O/C'd im sorta wondering if the processor or mobo is fucked. IDK though b/c I have never heard of a PC failing the benchmarks b4 that wasn't o/c'd.
I run prime 95 at about 50 degrees C with my barton 2500+
"1 Gig Centron DDR400 PC2700 RAM"
This doesn't make sense btw, DDR 400 is PC 3200...
Make sure you have a good look at the bios settings dude ;)


PS -make sure you only change one setting at a time in bios regarding your fsb or multiplier, and don't make it a big change just small increments.
Edit:
I have heard prime 95's torture test is the toughest bench out there

Yeah, it's the retail stuff. I'll go grab a new HSF and see what happens.

I bumped the core up to 1.675, and ran Prime95 again. Failed at the same old spot it always does. I watched the voltage and it started out fine at 1.65 to 1.66, and then about 2 minutes into the test it dropped back down to 1.6 to 1.62. Then if failed.

Right now after the test my CPU is running at 43.5C and the case is a balmy 25C


So you're leaning more towards I'm running into some sort of heat issue? Or is this just something you think needs to be rectified? I'm really wondering about the stability of the mobo at this point, or am I just being paranoid?
I'm still leaning towards the voltage. Increase it to 1.7 and see what happens, but again, ONLY do it after you have reseated your HSF and applied thermal paste (don't use the shitty pad). My case is 27C and under full load I never exceed 55C.

U\I know what AMD says, but if my 2600+ hits 55 or above, it usually slows down to a crawl or crashes. Below that it is fine though.

"1 Gig Centron DDR400 PC2700 RAM"
This doesn't make sense btw, DDR 400 is PC 3200...
Make sure you have a good look at the bios settings dude ;)



DOH! You caught the mistake! I'll have to kill you now.

I ended up getting called into work last night, so I didn't get a chance to go anywhere to grab a better HSF and some paste. Since todays Easter, I'm pretty much fucked til tomorrow. I'm doing the old box fan to the side of the case trick right now, and running benches. Low and behold...Prime95 will pass 2 out of 3 times. I may replace the mobo just in case.

Thanks for all the help guys, I'll let you know how it turns out.

don't get a new board, the problem is cooling dude. Drop the clock on the processor until you get a new fan(s) and heat sink.

I just got back from Microcenter, and picked up the following:

Cooler Master HSF
Thermaltake Video Card Cool
Artic Silver Thermal spooge
North and South bridge HSFs
1 80mm Sunon for my soon to be blowhole

I'll be tearing apart my machine shortly, but I wanted to see if you guys had any other ideas.

After applying the silver spooge and the new CoolerMaster HSF, the highest I can get my CPU is 50.5C and my case is a balmy 24C I can pass all the prime95 tests ( well, for an hour, haven't tested it more than that ) and I can play UT for more than a map or two.

The vid card cooler I got doesn't like my AiW, and I'm scared as shit of trying to get these push pins out for the northbridge cooler, so I'll probably take them back and pick up Painkiller or something.


Thanks for the help guys. I was under the impression I was well within spec for the heat, but apparently not. Now I know better. :)

yeah vid cards like the 9800 and the 5900 come with pretty fuckin good coolers to begin with, and are hard to beat with aftermarket stuff. I am considering water cooling the processor but I don't see any benefit at all with doing my vid card. The vid card heatsinks that cover the whole printed circuit board and the gpu are probably better than going with aftermarket stuff I figure.
Some things are just better left alone. I wouldn't bother with the north and south chips either unless I was really o/cing.

Sounds like thermal protection circuitry on your motherboard was throttling the cpu.